Thermal Characterization of a Tier0 Datacenter Room in Normal and Thermal Emergency Conditions

Datacenters are at the heart of AI, Industry 4.0 and cloud revolution. A datacenter contains a large number computing nodes hosted in temperature-controlled room. Due to increasing total power density nodes, overall compute capacity is often capped by peak consumption temperature bottlenecks. To preserve homogeneous performance assumption between all complex cooling solution required, but they might not be sufficient. In this work, we analysed characterised thermal properties Tier0 deploying advanced hybrid technologies: specifically, studied spatial temporal heterogeneity during production emergency hazards. This paper gives first quantitative evidence bottlenecks real-life workload, showing presence significant which could exploited thermal-aware job scheduling datacenter-room run-time workload adaptation distribution.